IntroductionPer- and polyfluoroalkyl substances (PFAS) are environmentally persistent, potentially carcinogenic chemicals. Previous studies investigating PFAS exposure and prostate cancer yielded mixed findings. We aimed to investigate associations between PFAS exposure and incident prostate cancer in a large cohort of U.S. men, overall and by selected demographic, lifestyle, and medical-related characteristics. MethodsWe conducted a case-cohort study among Cancer Prevention Study-II LifeLink Cohort participants who, at baseline (1998–2001), had serum specimens collected and no prior cancer diagnosis. The study included all men diagnosed with prostate cancer (n = 1610) during follow-up (baseline–June 30, 2015) and a random sub-cohort of 500 men. PFAS concentrations [perfluorohexane sulfonic acid (PFHxS), perfluorooctane sulfonate (PFOS), perfluorononanoic acid (PFNA), and perfluorooctanoic acid (PFOA)] were measured in stored serum specimens. We used multivariable Cox proportional hazards models to estimate associations between PFAS concentrations and prostate cancer, overall and by selected characteristics (grade, stage, family history, age, education, smoking status, and alcohol consumption). ResultsProstate cancer hazards were slightly higher among men with concentrations in the highest (Q4) vs lowest quartile (Q1) for PFHxS [hazard ratio (HR) (95% CI): 1.18 (0.88–1.59)] and PFOS [HR (95% CI): 1.18 (0.89–1.58)], but not for PFNA or PFOA. However, we observed heterogeneous associations by age, family history of prostate cancer (PFHxS), alcohol consumption (PFHxS), and education (PFNA). For example, no meaningful associations were observed among men aged <70 years at serum collection, but among men aged ≥70 years, HRs (95% CIs) comparing Q4 to Q1 were PFHxS 1.54 (1.02–2.31) and PFOS 1.62 (1.08–2.44). No meaningful heterogeneity in associations were observed by tumor grade or stage. ConclusionsOur findings do not clearly support an association between the PFAS considered and prostate cancer. However, positive associations observed in some subgroups, and consistently positive associations observed for PFHxS warrant further investigation.